Sundance doc'Ghost in the Machine' draws a damning line between AI and eugenics The Sundance documentary boldly declares that the pursuit of artificial intelligence, and Silicon Valley itself, is rooted in eugenics. Director Valerie Veatch makes the case that the rise of techno-fascism from the likes of Elon Musk and Peter Thiel is a feature, not a bug. That may sound hyperbolic, but, which is built around interviews with philosophers, AI researchers, historians and computer scientists, leaves little room for doubt. If you've been following the meteoric rise of AI, or Silicon Valley in general, Veatch's methodical deconstruction of the technology doesn't really unearth anything new. The film begins with the utter failure of Microsoft's Tay chatbot, which wasted no time in becoming a Hitler-loving white supremacist .

In a world where ICE agents are shooting US citizens on the street, the need for militias and extremist groups like the Proud Boys to support far-right interests has evaporated. Whether it was protesting Covid lockdowns, attending school board meetings, or facing off against Black Lives Matter protesters, the far-right Proud Boys were always on hand to support Donald Trump's first term in office. When Trump left office in 2021, the group's leaders languished in jail for their role in the January 6 attack on the Capitol. With reported infighting destabilizing the movement, it looked like the group's glory days were behind it. But Trump's return a year ago, and his release of all January 6 prisoners, signaled that a Proud Boy comeback could be in the cards.

Trump Warned of a Tren de Aragua'Invasion.' Hundreds of records obtained by WIRED show thin intelligence on the Venezuelan gang in the United States, describing fragmented, low-level crime rather than a coordinated terrorist threat. Alleged members of Tren de Aragua sit handcuffed during a preliminary hearing on July 9, 2025, in Santiago, Chile, where they faced homicide charges. As the Trump administration publicly cast Venezuela's Tren de Aragua (TdA) as a unified terrorist force tied to President Nicolás Maduro and operating inside the United States, hundreds of internal US government records obtained by WIRED tell a far less certain story. Intelligence taskings, law-enforcement bulletins, and drug-task-force assessments show that agencies spent much of 2025 struggling to determine whether TdA even functioned as an organized entity in the US at all--let alone as a coordinated national security threat.

'Extremist movements are using voice-generating bots to recreate the voices and speeches of major figures in their milieu.' 'Extremist movements are using voice-generating bots to recreate the voices and speeches of major figures in their milieu.' Extremists are using AI voice cloning to supercharge propaganda. Experts say it's helping them grow W hile the artificial intelligence boom is upending sections of the music industry, voice generating bots are also becoming a boon to another unlikely corner of the internet: extremist movements that are using them to recreate the voices and speeches of major figures in their milieu, and experts say it is helping them grow. "The adoption of AI-enabled translation by terrorists and extremists marks a significant evolution in digital propaganda strategies," said Lucas Webber, a senior threat intelligence analyst at Tech Against Terrorism and a research fellow at the Soufan Center.

The potential for rapidly - evolving frontier artificial intelligence (AI) models - especially large language models (LLM s) - to facilitate bioterrorism or access to biological weapons has generated significant policy, academic, and public concern. Both model developers and policymakers seek to quantify and mitigate that risk, with an important element of such efforts being t he development of model benchmarks that can assess the biosecurity risk posed by a particular model. This paper describes the first component of a novel Biothreat Benchmark Generation (BBG) Framework . The BBG is designed to help model developers and evalua tors reliably measure and assess the biosecurity risk uplift and general harm potential of existing and future AI models, while accounting for key aspects of the threat itself that are often overlooked in other benchmarking efforts, including different act or capability levels, and operational (in addition to purely technical) risk factors. To accomplish this, the BBG is built upon a hierarchical structure of biothreat categories, elements and tasks, which then serves as the basis for the development of task - aligned queries. As a pilot, the BBG is first being developed to address bacterial biological threats only. This paper outlines the development of this biothreat task - query architecture, which we have named the Bacterial Biothreat Schema, while future papers will describe follow - on efforts to turn queries into model prompts, as well as metrics for determining the diagnosticity of these prompts for use as benchmarks and how the resulting benchmarks can be implemented for model evaluation. Ov erall, the BBG F ramework, including the Bacterial Biothreat Schema, seek to offer a robust, re - usable structure for evaluating bacterial biological risks arising from LLMs, a structure that allows for multiple levels of aggregation, captures the full scope of technical and operational requirements for biological adversari es, and accounts for a wide spectrum of biological adversary capabilities.

The 30-year-old Virginia resident evaded capture for years after authorities discovered pipe bombs planted near buildings in Washington, DC, the day before the January 6, 2021, Capitol attack. Prince William County police seal the street in front of the home of suspected January 6, 2021, pipe bomber on December 4, 2025, in Woodbridge, Virginia. Federal agents have arrested a suspect identified as Brian Cole. Federal agents on Thursday announced the arrest of a suspect charged with planting the two pipe bombs discovered near the US Capitol complex on the eve of January 6, 2021 . Authorities identified the man as Brian J. Cole Jr., a resident of Woodbridge, Virginia.

This material may not be published, broadcast, rewritten, or redistributed. Quotes displayed in real-time or delayed by at least 15 minutes. Market data provided by Factset . Powered and implemented by FactSet Digital Solutions . Mutual Fund and ETF data provided by Refinitiv Lipper .

Safety evaluation of multimodal foundation models often treats vision and language inputs separately, missing risks from joint interpretation where benign content becomes harmful in combination. Existing approaches also fail to distinguish clearly unsafe content from borderline cases, leading to problematic over-blocking or under-refusal of genuinely harmful content. We present Vision Language Safety Understanding (VLSU), a comprehensive framework to systematically evaluate multimodal safety through fine-grained severity classification and combinatorial analysis across 17 distinct safety patterns. Using a multi-stage pipeline with real-world images and human annotation, we construct a large-scale benchmark of 8,187 samples spanning 15 harm categories. Our evaluation of eleven state-of-the-art models reveals systematic joint understanding failures: while models achieve 90%-plus accuracy on clear unimodal safety signals, performance degrades substantially to 20-55% when joint image-text reasoning is required to determine the safety label. Most critically, 34% of errors in joint image-text safety classification occur despite correct classification of the individual modalities, further demonstrating absent compositional reasoning capabilities. Additionally, we find that models struggle to balance refusing unsafe content while still responding to borderline cases that deserve engagement. For example, we find that instruction framing can reduce the over-blocking rate on borderline content from 62.4% to 10.4% in Gemini-1.5, but only at the cost of under-refusing on unsafe content with refusal rate dropping from 90.8% to 53.9%. Overall, our framework exposes weaknesses in joint image-text understanding and alignment gaps in current models, and provides a critical test bed to enable the next milestones in research on robust vision-language safety.

Recent research on large language model (LLM) jailbreaks has primarily focused on techniques that bypass safety mechanisms to elicit overtly harmful outputs. However, such efforts often overlook attacks that exploit the model's capacity for abstract generalization, creating a critical blind spot in current alignment strategies. This gap enables adversaries to induce objectionable content by subtly manipulating the implicit social values embedded in model outputs. In this paper, we introduce MICM, a novel, model-agnostic jailbreak method that targets the aggregate value structure reflected in LLM responses. Drawing on conceptual morphology theory, MICM encodes specific configurations of nuanced concepts into a fixed prompt template through a predefined set of phrases. These phrases act as conceptual triggers, steering model outputs toward a specific value stance without triggering conventional safety filters. We evaluate MICM across five advanced LLMs, including GPT-4o, Deepseek-R1, and Qwen3-8B. Experimental results show that MICM consistently outperforms state-of-the-art jailbreak techniques, achieving high success rates with minimal rejection. Our findings reveal a critical vulnerability in commercial LLMs: their safety mechanisms remain susceptible to covert manipulation of underlying value alignment.

Recent progress in video generative models has enabled the creation of high-quality videos from multimodal prompts that combine text and images. While these systems offer enhanced controllability, they also introduce new safety risks, as harmful content can emerge from individual modalities or their interaction. Existing safety methods are often text-only, require prior knowledge of the risk category, or operate as post-generation auditors, struggling to proactively mitigate such compositional, multimodal risks. To address this challenge, we present ConceptGuard, a unified safeguard framework for proactively detecting and mitigating unsafe semantics in multimodal video generation. ConceptGuard operates in two stages: First, a contrastive detection module identifies latent safety risks by projecting fused image-text inputs into a structured concept space; Second, a semantic suppression mechanism steers the generative process away from unsafe concepts by intervening in the prompt's multimodal conditioning. To support the development and rigorous evaluation of this framework, we introduce two novel benchmarks: ConceptRisk, a large-scale dataset for training on multimodal risks, and T2VSafetyBench-TI2V, the first benchmark adapted from T2VSafetyBench for the Text-and-Image-to-Video (TI2V) safety setting. Comprehensive experiments on both benchmarks show that ConceptGuard consistently outperforms existing baselines, achieving state-of-the-art results in both risk detection and safe video generation. Our code is available at https://github.com/Ruize-Ma/ConceptGuard.